Nurse scheduling problem (NSP) is a NP-hard problem. This report presents a parallel varied-length genetic algorithm for quality improvement of the nurse scheduling. Performance evaluation is based on the data collected from an area teaching hospital in central Taiwan and the results of this approach are compared with other heuristic algorithms such as a constraint genetic algorithm, simulated annealing, and tabu search etc. Computational results show that the parallel varied-length genetic algorithm integrated with an island model outperforms other methods.